Neural networks technique based signal-from-background separation and design optimization for a W/quartz fiber calorimeter

We present a signal-from-background separation study based on neural networks technique applied to a W/quartz fiber calorimeter. Performance results in terms of signal efficiency and improvement of the signal-to-background ratio are presented. We conclude that by using neural networks we can efficiently separate signal from background and achieve a signal enhancement over the background of the order of several thousands at high efficiency.